1
Chapter 5: Case Studies and Comparative History
None of the inventories of the causes of democracy surveyed in Chapter 4 can be taken at
face value because they summarize conclusions from disparate and poorly integrated sources of
uneven rigor. Before giving credence to checklist items, it is essential to find out where they came
from. The remainder of this book therefore evaluates the strengths and weaknesses of the methods
and approaches employed by the scholars who have launched these propositions into our collective
academic mind.
If there were a way to measure quantities of information there would be no doubt that the
bulk of our knowledge of democratization comes from histories and case studies. It is ironic that
non-comparative research so dominates a field known as “comparative” politics, but this is an irony
that is longstanding and well known. The huge number of books and articles about the birth, death,
or survival of democracy in dozens of countries in the twentieth century alone is more than any one
scholar can digest. In fact, anyone unlucky enough to be buried under this small mountain of printed
matter would surely be suffocated and crushed. To be sure, this literature does not satisfy all the
criteria for good theory. I will argue in this chapter that histories and case studies produce a kind of
knowledge that lacks generality and theoretical integration. However, this kind of knowledge is
unsurpassed in its thickness; hence its bulk. In this chapter I will also survey and critique a smaller
but still sizable body of literature that consists of comparative histories: detailed comparisons of the
historical development of a (usually) small number of cases. Comparative history reaches out toward
generality and theoretical integration, but still has enough characteristics in common with case
studies and histories to warrant discussing them jointly.
2
Histories and Case Studies1
Histories (written by historians) and case studies (written by political scientists) are different
genres with zealously policed boundaries. In fact, one of my purposes in discussing them together is
to highlight their differences. However, they also have two advantages in common. First, they are
thick: multifaceted, detailed, conceptually rich, and multidimensional. (See Chapter 3 for a fuller
definition of thickness.) Second, they analyze change over time in a way that makes especially solid
causal inferences possible.
Histories and case studies are the best examples of a "thick" approach. Both paint
multifaceted, multidimensional portraits of countries and do so in rich detail. Studies of
democratization in single countries, for example, are so richly detailed that they do not attempt to
explain anything as grand as "democratization." Instead, they explain a series of specific events that,
taken together, amount to democratization. A typical history is R.L. Webb's Modern England (Webb
1980). England was clearly not a democracy at the book's opening in 1760 but had clearly become a
(male) democracy 420 pages later in 1918. Yet no one event marks the democratization of England.
Rather, democratization is the accumulation of small reforms: the separation of the bureaucracy
from the King's household, the development of a prime minister as a first among equals, the
toleration of other religions, the formation of principled political parties, the development of the
principle of collective responsibility, the publication of parliamentary debates, the obligation of the
monarch to appoint the ministers chosen by parliament, the elimination of the royal veto,
elimination of prior censorship of the press, the equal apportionment of electoral districts, the
enfranchisements of middle- and working-class men, the responsibility of the cabinet to the
Commons, and the diminution of the legislative powers of the House of Lords. Webb's story of
these reforms makes them impossible to understand apart from the personalities of monarchs and
3
politicians, personal rivalries and reputations, deaths of major figures, and the formation and
dissolution of cabinets, all played against the backdrop of wars, economic crises, economic
transformation and social and technological change.
Similarly, a case study of the consolidation of democracy in Venezuela, Daniel Levine's
Crisis and Political Change in Venezuela, recounts in detail crucial events of the first, short-lived,
democratic regime of 1945-1948 and the first decade of the second democratic regime that began in
1958 (Levine 1976). It also analyzes the formation of party factions and the origins of party splits,
explains negotiations over education and social policy, and characterizes the personalities of several
key political leaders. Like a history, it reconstructs events, focuses attention on decisions made by
individuals, and situates micro-level processes within macro-level institutional, economic, social, and
international contexts. Both Webb's and Levine's books, and the histories and case studies they
represent, deserve to be called "thick" in the best sense of the term.
Because they are thick and analyze change over time, histories and case studies share the
ability to support powerful causal inferences. Causal inference--attributing causation of an outcome
to some stimulus or stimuli--is hard. In fact, as Popper argued, we can never prove that a causal
relationship exists; the best we can do is to disprove other hypothetical causal relationships: an
indirect "proof," at best (Popper 1968). What makes causal inference particularly hard in political
science is the complexity of the political world. Recall the argument from Chapter 3 that politics is
complex; that events have many facets and that outcomes are the product of a long and densely
tangled chain of causation. This complex nature urges us to build complex, multifaceted
explanations. But the more elements there are to an explanation, the greater the danger that some of
them will be spurious or false, and the more difficult it becomes to disconfirm all the possible
explanations except the one we wish to "prove," indirectly. (Chapter 7 will develop this argument
4
further.)
Indirectly "proving" an argument about causes in politics therefore necessarily involves
holding constant, or controlling for, as many alternative hypothetical causes as possible. Histories
and case studies excel at this because they take advantage of change over time in a single case. In
effect, what they do is compare each country to itself at an earlier time. This is a far more foolproof
method than comparing one country to another, because every country has much more in common
with itself in the recent past than it does with any other country at any time. The logic of inference is
therefore that any feature that did not change from time 1 to time 2 cannot be a cause of an outcome
observed at time 2. Because most features of a country do not change quickly, these static features
can be ruled out as causes by this method and analysts can more confidently attribute causal force to
the features that did change just before the outcome. The shorter the intervals between observations
and the more finely the process is observed, the greater the confidence the inference inspires. A
sweeping history of several centuries in one chapter probably inspires little confidence, but tracking
events decade by decade begins to inspire confidence. Tracking them year by year or month by
month is a vast improvement, and studies that manage to reconstruct events week by week or day by
day begin to seem irrefutable. It becomes difficult to imagine an alternative story that would be
consistent with all the details.
It is this historical dimension that grants histories and case studies of democratization their
authority. For example, Levine's claim that moderating leadership helped consolidate Venezuelan
democracy was bolstered by his accounts of attempts at radical mobilization that moderate party
elites managed to isolate and defuse or marginalize (Levine 1976: ch. 5). Similarly, Webb can show
that the Glorious Revolution of 1688, installing William and Mary as the monarchs of England,
established the supremacy of Parliament over the King (Webb 1980). This kind of historical process-
5
tracing also enabled Alfred Stepan to show that President João Goulart's decision to mobilize
landless peasants provoked Brazilian elites to conspire to overthrow him and that newspaper
editorials helped legitimate the conspiracy (Stepan 1971:191-204). Comparisons within countries
over time are the most powerful approach at our disposal for identifying the immediate causes of
specific events.
Of course, there are some important differences between histories and case studies. Two
complementary differences come to mind. The first is simply a matter of degree: most histories are
thicker than most case studies in political science. There are exceptions, such as Jared Diamond's
sweeping history of humanity, Guns, Germs, and Steel, which is necessarily thinner due to its scope.
And case studies can be unusually thick, such as the Graham Allison classic Essence of Decision, which
reconstructs decision-making during the Cuban Missile Crisis day by day, sometimes hour by hour
(Allison 1971; Diamond 1997). Nevertheless, the general tendency is for histories, especially
histories written by academics for academics, to be thicker than political science case studies. The
latter may go into fine detail at points, but tend to leap from crucial event to crucial event, and not
necessarily in chronological order. They do not sustain a continuous, detailed narrative.
This is not the only difference between these two genres, however. If it were, we would have
to conclude that case studies are merely inferior histories. As Juan Linz once remarked to a
convention audience, "If we do our work very well, we are almost historians."2 The other difference
is that case studies in comparative politics are more narrowly and explicitly focused on explanation.
Political science case studies almost always explicitly identify the independent variables and put the
causal process front and center. They are organized to advance and support an argument about what
caused what. They also, whenever possible, embed their arguments in a larger theoretical framework.
They self-consciously present the case as a specific instance of more general rules. Case studies are,
6
by design, installments in the larger enterprise of theory-building. Histories also seek to explain, but
it is just one of their purposes, which include recounting the facts completely and accurately,
interpreting what events meant to people at the time, and simply telling a good story. Historians vary
in their mix of purposes. Some emphasize the scholarly recounting, some tell better stories, and
some are more didactic than others. But none are as single-mindedly focused on demonstrating that
the outcome was a necessary consequence of the explanatory factors. Histories generously grant more
of a role to chance, whim, accidents, mistakes, coincidences, and miscalculations, which make for a
more colorful story. These are exactly the elements that render a theoretical explanation logically
incomplete (see Chapter 3). As E.P. Thompson observed (even while defending the notion of a
“logic of history”),
Historical concepts and rules. . . display extreme elasticity and allow for great irregularity; the
historian appears to be evading rigour as he disappears into the largest generalisations at one
moment, while at the next moment he disappears into the particularities of the qualifications
in any special case. This provokes distrust, and even laughter, within other disciplines
(Thompson 2001:454).
Judged by the criteria of political science, histories are merely prolix, equivocating case studies.
By stereotyping histories as inferior case studies and case studies as inferior histories, I do
not mean to imply that one is superior to the other; I merely mean to show that virtue in one
discipline often appears to be a vice in another. The same is true for the vices that histories and case
studies share when judged by the standards of other approaches in comparative politics. There are
three shared vices in particular: myopia, an inability to generalize, and a tendency to capitalize on
chance.
Myopia–already mentioned in Chapter 4–is the tendency to exaggerate the impact of short-
7
term, micro causes. This tendency is practically inevitable in any approach that involves close
examination of cases. Mom was right: squinting causes near-sightedness.3 Looking at events in
historical perspective, especially over closely-spaced intervals of time, necessarily privileges causal
factors that change over those intervals and discredits causal factors that remain unchanged
throughout the period of observation. Case studies of democratization therefore tend to emphasize
what is dynamic in the short term: leadership, natural disasters, economic crises, and wars. This
tendency has already been noted, in Chapter 4, in the critique of inductive elite theories of
democratization, but examples can be found in most case studies.
One good example is Malloy and Gamarra’s Revolution and Reaction: Bolivia, 1964-1985, which
analyzes Bolivian politics during the period from the breakdown of the revolutionary regime to the
tortuous democratic transition of 1980-1982 (1988:84-228). Although the authors (both political
scientists) make it clear that chronic patrimonialism, corruption, human rights abuses, and an
unsustainable development strategy undermined support for every government during this period,
the factors that drive the narrative forward, from coup to election to coup to election, are specific
short-term events: a hunger strike, a party split, an uprising by one or another military faction,
pressure from the Carter administration, the failure of negotiations with Chile over an outlet to the
Pacific Ocean, the power-hungry personality of Gen. Hugo Bánzer, and popular reactions to
repression by the García Meza government. The authors conclude that “if [democracy] comes, it will
be produced by serendipitous salidas [exit pacts] reflecting the pragmatic creativity of political elites
and not grand solutions reflecting intellectual imports or architectural schemes” (Malloy and
Gamarra 1988:225).
Another example–this time from a relatively stable case–is González and Gillespie’s analysis
of the impact of presidentialism on regimes in Uruguay (González and Gillespie 1994). Their goal is
8
to attribute regime crises to the presidential features of the Uruguayan constitution, but the task is
quite a challenge because Uruguay suffered only one coup (1973) and one semi-coup (1933) from
1918 to 1994 and presidentialism was a constant during that time. If a constant feature such as
presidentialism caused breakdowns only twice in 77 years, then the explanation must be incomplete.
Something else much explain why coups happened in those years and not others. González and
Gillespie therefore fill in the gaps with short-term, specific events and actions: the death of President
Oscar Gestido in 1967, his replacement by the more hard-line President Jorge Pacheco, a Tupamaro
guerrilla offensive, repression of leftists, the loss of a governing majority in congress, Pacheco’s
selection of an unpopular successor, a pact between the two largest parties to support a crackdown,
accusations of corruption, and the political isolation of the president (González and Gillespie
1994:233-7). If the authors were not determined to blame presidentialism, it probably would not
figure in the explanation at all.4
In neither of these examples, nor in case studies generally, do scholars ignore or completely
discount structural causes such as social structure, geography, constitutional design, or culture.
Myopia is not a failure to see large, distant objects at all, but an inability to see them in focus or,
mixing metaphors, to give them their proper weight. Case studies and histories tend to give too
much weight to the dynamic particulars they can portray clearly and too little weight to the large,
immobile features of the landscape in the background that they take for granted.
The specificity of these approaches also interferes with generalization. The difficulties for
generalization are conceptual, theoretical, and empirical. Conceptually, if one were to try to
generalize about democratization from the case of England, for example, it would first be necessary
to translate the particularities of the English process into concepts that usefully describe processes in
other countries. Some of the small reforms listed at the beginning of this chapter do not match up
9
easily with steps toward democracy in other countries. In particular, reforms that weakened the
monarchy are irrelevant for countries that never had a monarch. One of the reforms that has the
most traveling potential is the establishment of the responsibility of the cabinet to the Commons. If
we substitute “lower or sole chamber of the national legislature” for “Commons,” we have a reform
that is comparable to similar reforms in most other parliamentary democracies. However, it is still
irrelevant for the many presidential democracies, which permit the legislature to dismiss the
executive only in extraordinary situations. If we hold fast to the exact concepts used in case studies
and histories, it becomes difficult even to claim that the concept of “democracy” has any useful and
cross-nationally comparable meaning.
Theoretical specificity also hinders generalization. If concepts are too specific to travel,
theories using those concepts do not travel well, either. But even a theory expressed entirely in
generally-comparable concepts can be too specific for generalization if it combines a large number
of propositions. Imagine, for example, a theory that predicts that democratic regimes will survive in
countries that have a high standard of living, a small number of political parties, effective
representation of business interests, have never been colonies, and are found on islands. Each
proposition narrows the applicable domain of the hypothesis to an ever-more limited set of cases.
Many countries could satisfy the first three propositions, but ruling out former colonies excludes the
U.S., Australia, Canada, Iceland, and many others. Limiting the domain to islands shrinks the
domain to Japan and Britain; and if we tack on the additional requirement of a homogeneous
national identity, then only Japan is left! This happens even though all of the propositions are
expressed in terms of general concepts.
General empirical confirmation is more difficult for the hypotheses that arise from case
studies and histories because these studies involve implicit conceptual or theoretical specificity. The
10
authors of the case studies may never have intended for their propositions to be generalized, but
other scholars who try to generalize them find that the same propositions do not “work” as well
when applied outside of their original context. For example, many scholars have noted that the
survival of democracy in India is a puzzle because democracy rarely survives in other poor societies
with deep ethno-religious-linguistic cleavages. To explain this paradox, scholars have repeatedly
given some credit to British colonial rule, arguing that the British educated India’s elite in its liberal
values, created an efficient civil service, and gave Indians experience with limited self-government
(Weiner 1987, 20; Lipset 1993, 5). Even scholars who are more critical of the British role argue that
opposition to the British united Indians behind the Indian National Congress, which then
neutralized many social cleavages for decades after Independence, enabling democracy to survive.
Although some version of this hypothesis works well for India and also seems important for
understanding the survival of democracy in the United States, Canada, Australia, New Zealand, and
the English-speaking Caribbean, it does not help explain the failure of democracy in some other
former British colonies such as Nigeria, Pakistan, Gambia, Zimbabwe, Ghana, Malaysia, Singapore,
Kenya, and Zambia. The problem is conceptual: “British colonial rule” meant different things in
different colonies. At a minimum, we must distinguish between colonies of settlement (as in North
America and Australia) and colonies of conquest. There were also differences in the timing and
length of colonization, degrees of self-government and exploitation, and the nature of the
independence struggle that may help resolve the Indian paradox. Applying watered-down versions
of a proposition without its original contextual qualifiers amounts to “concept-stretching” (Sartori
1970). We should not be surprised that hypotheses built from stretched concepts do not hold up
when taken out of context.
Another illustration comes from case studies of oil producers arguing that dependence on oil
11
exports either prevents democratization or undermines any democratic regime that comes into
being.5 This proposition gains support from non-democracies in the Middle East and from fragile
democracies such as Venezuela and Ecuador, but it founders on the shoals of Britain and Norway.
One could conclude from this that the proposition is false, but it is more likely the case that it is true
in some contexts but not others. Terry Karl has argued, for example, that dependence on oil exports
is antithetical to democracy only where the oil industry developed before a bureaucratic state
became consolidated; in Britain and Norway, an efficient state bureaucracy (and democratic regime)
developed long before North Sea oil was discovered (Karl 1997). Karl’s argument in effect made an
implicitly specific theory explicitly specific, which made it clearer that it had a limited domain.
Propositions generated by case studies tend not to survive attempts at generalization because they
are, in reality, conditional on other theoretical propositions that are taken for granted by scholars
focusing on a single case, and therefore left implicit.
The three problems of generalization just described are all problems for formulating theories
or hypotheses that are likely to be generally true. Generalization, however, also concerns inference:
drawing general conclusions from limited evidence. Case studies and histories are also handicapped
for this task, for one can never infer that a proposition is generally true because it is true in one case.
Nevertheless, there have been attempts to argue that although case studies cannot confirm a
proposition, they can disconfirm one. In 1975 Harry Eckstein argued that a single case could
disconfirm a general hypothesis if it was a “crucial case”: one in which the hypothesis must hold true
if it is true for any case (Eckstein 1975). More recently, Douglas Dion used Bayesian conditional
probability theory to show that some hypotheses about necessary conditions could be generally
disconfirmed if they were falsified in a small number of cases, even a single case. However, this is
true only when one already knows that the condition of interest probably is necessary and that any
12
alternative explanations probably are not true (Dion 1998). In other words, a single case is unlikely
to tell us something we do not already know. Therefore, Lieberson’s conclusions about the limits of
small-sample testing still apply with special force to case studies: “Put bluntly, application of Mill’s
methods to small-N situations does not allow for probabilistic theories, interaction effects,
measurement errors, or even the presence of more than one cause” (Lieberson 1992). Actually,
Eckstein admitted as much in 1975; he merely asserted that adequately measurable, deterministic,
monocausal relationships were not so rare in comparative politics.
In all subsequent chapters of this book, I will summarize the findings about democratization
uncovered by each approach. I will not do so with respect to histories and case studies, however, for
the reason I have already given: they do not lend themselves to generalization. The conclusions of
case studies of democratization in Britain, Ghana, Russia, Taiwan, Peru, or any other single case
usually were not intended to be stated out of context. And even if some authors misguidedly
presented them as general lessons, we should not interpret them as such because they could be taken
as true only under implicit conditions that need to be made explicit. None of this should detract
from the worth of case studies which, again, are the sources of our most rigorous explanations of
the immediate causes of specific events. If you want to understand the birth or death of democracy
in one country, read a case study. Better yet, read a history. But no one is now in a position to
summarize the conclusions of these studies along with all the contextual conditions attached to each
conclusion that are needed to avoid overgeneralization. If someone were in a position to do this, we
would have a better grasp of the kind of general theory toward which we strive: a kind of theory that
becomes more complex, not less, as it becomes more general.
The third vice that case studies and histories share–in addition to myopia and an inability to
generalize–is a tendency to “capitalize on chance,” i.e., to confuse mere coincidences and spurious
13
relationships with causal relationships. This kind of confusion is always a danger when alleged causal
relationships are not tested with evidence different from the evidence that originally suggested the
relationship. A good illustration is the claim made by some specialists on U.S. politics that the
ingenious design of the U.S. constitution is responsible for the longstanding success of democracy in
the United States (Calabresi 2001; Diamond 1959). As long as attention is confined to the United
States, it is very hard to challenge this notion, especially when it is easy to come up with examples of
potentially regime-threatening conflicts that were resolved according to constitutional procedures.
One way to rule out coincidences would be to carry out a truly comparative analysis. Including Latin
American cases in the study would quickly cast doubt on the importance of the U.S. constitution,
because most Latin American constitutions since the 1820s have incorporated most of its major
features–presidentialism, bicameral legislature, supreme court, and in many cases federalism as
well–yet Latin America is known for the instability of its democratic regimes. The constitution alone
is clearly not a sufficient explanation (Riggs 1988).
Different evidence need not come from other countries. In principle, a scholar could base a
hypothesis on one set of observations from one country and then test it with a different set of
observations from the same country (King et al. 1994, 223-8). There are a great many within-case
research designs, which the field of American politics exploits to the hilt–surveys, comparisons of
states, analysis of time series, and so on–and all of these are available for case studies in comparative
politics as well. Still, an insuperable problem arises when the outcome and its alleged cause are
invariant. In order to test a hypothesis, the cause and the outcome must vary. This is the only way to
tell whether they covary, and covariance is an essential attribute of causal relationships. If changes in
Y do not correspond to changes in X, we can rule out a causal relationship; but if X and Y do not
change, we can never be sure. Maybe Y would have happened without X; maybe Y could change
14
even if X did not. Until X or Y changes, these possibilities are just hypothetical guesses. We can tell
different stories about their relationship–spin different theories–but it all remains hypothetical until
we can observe actual change.6
Because every country has many relatively constant, unchanging features, and because
academics are highly intelligent and creative people who can dream up causal connections among
them without breaking a sweat, case studies and histories tend to identify ostensibly important
invariant “causes” without providing us with any means for judging their relative importance, or
whether they are important at all. Why, for example, has Britain been a democracy without
interruption for so long? Is it because it is an island separate from the European mainland, which
lessened the need for a standing army that might also have been used to repress its own citizens? Is
it because the English variety of feudalism and the early establishment of a limited monarchy created
traditions and cultural expectations that quickly defeat attempts to encroach on democratic rule? Is it
because England was the first country to industrialize? All of these characteristics were acquired so
far in the past that today they are constants, so as long as democratic survival remains a constant as
well, it is very difficult to show that they are or are not causes of British democracy. Other examples
of invariant causes are geography, climate, or “national character.” Today we rarely turn to them for
our explanations, but all these factors lead to the same problem.
The same problem arises in assessing explanations for the long-term absence of democracy in,
for example, Egypt. Maybe it is the lack of any prior experience with democracy; maybe it is that
Egypt has a majority Moslem society; maybe it is the Egyptian state’s reliance on oil exports; maybe
it is Egypt’s colonial past. Again, there is no way to be sure, because neither these possible causes
nor the lack of democracy vary in Egypt. Similar arguments have often been made about entire
regions, not just countries: democracy is associated with Western Europe’s feudal past, Christianity,
15
or being in the “core” of the world economy, while weak or absent democracy is associated with
non-European attributes such as Islam, Confucianism, or being in the “periphery” of the world
economy (Huntington 1993; Wallerstein 1974). One could even extend this kind of reasoning to
explain chronic regime instability in, for example, Peru. Democracy has rarely lasted long in Peru
because it was an important part of the Spanish colonial empire; because a deep ethnic and linguistic
cleavage divides the coast from the sierra; because a decades-long feud between the populist APRA
party and the armed forces has never been resolved; or because El Niño periodically wreaks havoc
with the economy and the environment.
Actually, hypotheses of this sort can be tested, but only if we have variation of some sort. It
could be variation among subnational regions, as in Putnam et al.’s Making Democracy Work (Putnam,
Leonardi, and Nanetti 1993); it could be variation over time, if a sufficiently large time span is
considered; or it could be variation among individuals, if their opinions are relevant for the
hypothesis and a survey can be done. It is also possible to break down the dependent variable to
expose its inner clockwork–the causal mechanism–so that we can observe how its components
move in synchrony. But these alternatives serve to reinforce the point that if cause and effect truly
are static and unchanging, their relationship cannot be tested unless and until new evidence is found.
What can we say, then, about the kind of knowledge that we find in histories and case
studies? First, they produce unparalleled descriptive knowledge in the form of concepts, facts, and
narratives. Within the discipline, description is valued less than causal inference, but this is unjust.
Good description is hard, it is an essential foundation for causal inference, and it is probably valued
more by those outside the discipline than our attempts at building causal models (Schedler 2003).
Furthermore, this descriptive knowledge tends to be custom-made for each case and therefore fits
exceptionally well. Second, histories and case studies also attempt to model causes, but do so with
16
three inherent limitations. Their causal arguments are biased in favor of short-term, micro-level
causes; the arguments are unlikely to translate easily to other times and places; and any static causes
remain at the level of hypotheses rather than tested propositions. Histories and case studies are great
ways to develop ideas about things that may matter generally, but cannot show that they do matter
generally.
The Nature of Comparative Histories
Comparative histories are intended to overcome the limitations of histories and case studies
while retaining some of their advantages. By analyzing and comparing a few cases in some detail,
comparative histories try to avoid myopia, move toward generalization, and test hypotheses against
fresh data. Compared to most other approaches, comparative history is a method that employs
relatively thick concepts and develops relatively thick theory, albeit not as thick as the concepts and
theories in case studies or histories. It tends to reject the goal of universality, but strives to generalize
within geographically and historically bounded domains. Comparative historians often claim to be
developing an integrated theory that is accumulating knowledge, as evidenced by a long tradition in
which successive researchers address and revise the conclusions of their predecessors. Indeed, the
roots of comparative history run so deep that it could be said to be the original approach of
comparative politics. Comparative historians rightfully claim Tocqueville, Marx, Weber, and
Durkheim as their ancestors. As structural-functionalism and behavioralism rose in the 1960s,
comparative history temporarily fell into disfavor. However, inspired by Barrington Moore’s The
Social Origins of Dictatorship and Democracy and joining in the reaction against structural-functionalism,
comparative history achieved a resurgence. Moore's thesis that privileged classes would maintain a
non-democratic regime unless they were swept away by a revolution was enthusiastically embraced
by many scholars in the radicalized intellectual environment of the late 1960s (Laitin forthcoming).
17
Theda Skocpol, Charles Ragin, and others developed methodological justification for the approach
in the 1980s, and by the 1990s comparative history became a more self-conscious and vigorous
approach (Evans, Rueschemeyer, and Skocpol 1985; Ragin 1987).
Taking stock of comparative historical analysis in 2003, James Mahoney and Dietrich
Rueschemeyer defined the approach by stating that "all work in this tradition. . . share[s] a concern
with causal analysis, an emphasis on processes over time, and the use of systematic and
contextualized comparison" (Mahoney 2003).7 This definition differentiates comparative history
from, respectively, descriptive histories, cross-sectional comparisons of static observations, and
efforts to develop universalistic theories. This definition characterizes well a series of books on
democratization and state-building in Western Europe that took Moore's Social Origins (1966) as its
starting point.8 The major works in this series include Skocpol's States and Social Revolutions (1979),
Luebbert's Liberalism, Fascism, or Social Democracy (1991), Downing's The Military Revolution and Political
Change (1992), and Ertman's Birth of the Leviathan (1997). This literature spread to Latin America with
Collier and Collier's Shaping the Political Arena (1991) and has since given rise to several important
books that transcend any single region and focus more narrowly on democratization: Rueschemeyer,
Stephens, and Stephens's Capitalist Development and Democracy (1992), and Ruth Collier's Paths Toward
Democracy (1999).
A superficial glance at the comparative historical literature suggests that it has developed in a
cumulative way, gradually elaborating and repeatedly testing a coherent theory of democratization
that applies to an expanding set of countries. Figure 5.1 sketches out the major regime types (or
stages of regime-building) found in the Western European countries that have been the principal
focus of this literature, and the time periods in which they predominated. The common
starting-point for all of them was the Holy Roman Empire, founded in 936 A.D. When the Holy
18
Roman Empire began to disintegrate in the 15th century, it left behind many small feudal states.
Some of these remained small feudal states for centuries; others were agglomerated into larger,
consolidated national states. This is the first major division explained in this literature. The second
bifurcation divided consolidated states that evolved into absolutist regimes in the 17th and 18th
centuries from those that preserved a rough balance between crown and nobility, the rights of towns
and guilds, and mutual obligations between landlords and peasants--a situation that Downing calls
"medieval constitutionalism." By the 19th and early 20th centuries, these states parted ways a third
time, either establishing the responsibility of the executive to parliament and expanding the suffrage
to become liberal democracies, or undergoing a social revolution. Many of the liberal democracies
lasted only a short time before they became social democracies (in Luebbert’s terminology) or fell
prey to fascism. Finally, democratic regimes either survived or gave way to authoritarian regimes. In
some countries there were repeated cycles of democracy and authoritarianism. The Latin American
countries covered in the works discussed here followed a different path until the 20th century. All of
them were originally European colonies, but when they became independent in the early 19th
century, their consolidation as national states was delayed by several decades of civil war or
dictatorship. However, the Latin American cases fit well into Figure 5.1 from about 1900 onward,
oscillating between democracy and authoritarianism.
The claim that comparative histories constitute a cumulative research program is bolstered
by their overlapping historical coverage. Figure 5.2 superimposes on the framework of Figure 5.1
the domains covered by the eight comparative histories discussed in this chapter. Downing (1992)
and Ertman (1997) cover approximately the same periods and propose answers to the same
questions (among others): How did the remnants of the Holy Roman Empire become consolidated
as states and why did some become absolutist regimes while others resisted absolutism? Moore
19
(1966) and Skocpol (1979), though published earlier, study the later political transformations of
either absolutism or consitutionalism into social revolution, fascism, or liberal democracy. Luebbert
(1991) focuses on transitions from constitutional or proto-democratic regimes to liberal democracy,
social democracy, or fascism, ignoring social revolution, which was missing in his exclusively inter-
war European sample. The Colliers (1991), in turn, omit social democracy and fascism (both missing
from their Latin American sample) to focus on transitions between authoritarianism and
democracy.9 The most sweeping study is that by Rueschemeyer, Stephens, and Stephens (1992),
which covers the gradual early democratization of Europe, recent transitions in Latin America and
the English-speaking Caribbean, and fascist interruptions of the early 20th century. Therefore, while
each study considers a different slice of history, collectively these comparative histories provide
continuous coverage of the last half-millenium in Western Europe and Latin America.
Despite the considerable overlaps, these works contain too many fundamental differences to
treat them as a fully integrated research program. Most fundamentally, they attempt to expalin
different outcomes. Obviously, only the works that extend into the years since about 1920 (when
most European countries extended the suffrage to all adult males) really attempt to account for what
we now consider democratic regimes. Moore and Skocpol barely reach this period, and only for
some of their cases. Downing and Ertman seek to explain only institutions or traditions that were
precursors to democracy, such as the balance between crown and nobility and the rights of towns
and peasants. But the differences in outcomes are not solely due to different historical foci. Even
authors dealing with the same period seek to explain different outcomes, and democracy is not
always the outcome of chief interest. While Downing is primarily concerned with representative
assemblies and other institutions that predisposed Europe toward democracy, Ertman is equally
interested in explaining the type of regime (constitutional or absolutist) and the type of state
20
(bureaucratic or patrimonial). Moore was interested in three different kinds of "revolutions": the
bourgeois revolution (allegedly leading to democracy), the "conservative revolution from above"
leading to authoritarianism, and the peasant revolution leading to communism. Skocpol was
primarily interested in explaining social revolutions; democracy was for her just one of several
possible outcomes in a residual category of non-revolution. Of all the scholars of more recent
periods, only Rueschemeyer, Stephens and Stephens are exclusively interested in explaining
democracy. Luebbert distinguished between liberal democracy and social democracy--economic
systems as well as political regimes--and contrasted both with fascism. The Colliers, in Shaping the
Democratic Arena, are ultimately interested in explaining democratic survival (as an expected outcome
of institutionalized party systems that incorporate labor in a peaceful way), but only in the final
chapters. Along the way, they spend considerable effort explaining prior outcomes--the
characteristics of parties, the mode of incorporation of labor, the nature of labor laws, and the
strength of labor and the oligarchy. In her later (1999) book, Ruth Collier is actually interested in
accounting for variations in the role of labor in democratization rather than for democratization per
se.
Comparative histories have also disagreed about which actors matter. Skocpol took Moore
to task for merely paying lip service to the notion of the state as an autonomous actor (Skocpol
1973). The same criticism could be made of Luebbert and perhaps Ruth Collier’s (1999) book.
Skocpol, Downing, Rueschemeyer et al., and Ertman emphasize foreign states as important actors;
Moore, Luebbert, and the Colliers do not. Similarly, authors disagree about the democratic
orientations of key actors. Moore (like most of his predecessors and contemporaries) associated
democracy with the interests of the bourgeoisie; if the bourgeoisie is defined to include the gentry
and the urban middle class, Downing and Ertman would agree. Luebbert argued that only some of
21
the middle class supported democracy, and only in some cases, as the outcome depended on the
middle class’s choice of allies. The central argument of Rueschemeyer, Stephens and Stephens,
however, is that the working class was a far more consistent friend of democracy than the middle
class, except in South America. This dispute was the inspiration for Ruth Collier’s book, which
identified seven different roles played by labor in democratization, some of which were supportive
and others, not.
Comparative histories have varied most with respect to the set of explanatory factors or
variables that they emphasize. The only factor shared by all of these works is the assumption that the
economic interests of social classes played an important role. Barrington Moore’s work is the purest
expression of this; the others add other variables to social class. As noted before, Skocpol added
state interests and international relations to the mix, initiating a tradition of centering comparative
historical analyses on class-class, class-state, and state-state relations (Mahoney 2003:151). Luebbert,
however, went beyond the class cleavage to consider pre-industrial religious, regional, linguistic, and
urban-rural cleavages. Downing emphasized, in addition to class, medieval institutions and
traditions, geography, military technology, trade, and colonial resources. The other works refer to a
still richer variety of explanatory factors, including commodity specialization, immigration,
leadership, modes of incorporation of labor, civil war, short-term economic performance,
transnational political ideas and polarization, the nature of the antecedent regime, and coalition
politics.
Although these works do not present a coherently integrated body of theory, they can be
said to share a “metatheory”–a looser set of assumptions–as James Mahoney has noted (Mahoney
2003:132-7). All comparative historical researchers make three fundamental assumptions, and a
subgroup makes a few more. One fundamental assumption is that big events such as revolution,
22
state-building, and democratization can be explained. Not all social scientists agree with this. Barbara
Geddes, for example, has argued that big events are too complex to explain and that we must focus
on the microfoundations of politics if we are to make any progress (Geddes 1997). A related
assumption is that big events have big (structural) causes. Most scholars who refer to "structures"
mean class structure, but other social structures such as religious or ethnic cleavages can be included,
and I would argue that other large-scale, slowly-changing features such as physical geography,
commodity specialization, and state capacity qualify as structures as well. Comparative historians do
not claim that all causes are structural, but their belief that some causes are structural sets them apart
from those who are exclusively interested in the effects of mutable institutions, culture, or strategic
calculations of elites.
A third fundamental assumption is that it is meaningful and useful to treat social classes as
actors. Some scholarly traditions long ago decided that "class" is not a useful concept except in the
rare cases where a social class is conscious, cohesive, and conspiratorial (Parry 1969: ch. 2).
Comparative historians tend to assume that members of classes do have a common economic
interest, that they are aware of their shared interests, and that they collectively pursue their common
interests. However, unlike primitive Marxians, comparative historians typically qualify these
assumptions in two respects. First, they tend to divide classes into more precise, and therefore more
homogeneous, categories. Instead of writing about capitalists in general, they distinguish among
agricultural landlords, industrialists, and finance capital, and often sub-categories of these. Instead
of the bourgeoisie, they distinguish among the rural middle class, urban professionals, merchants,
civil servants, and so on, often to the point of reducing classes to a set of occupations. Second,
comparative historians now tend not to assume that collective action in pursuit of common interests
comes naturally. Therefore, important supplementary actors in their analyses are class-based
23
organizations such as labor unions, political parties, and interest groups, which mobilize their
members in the defense of the interests that they define on behalf of the class. These two
modifications of the class-analytic approach do make it more useful. Taken to the extreme, it would
become indistinguishable from the interest-group politics approach, but most comparative histories
remain at a slightly more abstract level.
A sub-group of comparative historians shares several additional assumptions that help define
the approach. After a long intellectual evolution from the Marxian dogma that the state is merely "a
committee for managing the common affairs of the whole bourgeoisie," most contemporary
comparative historians now assert that states are autonomous actors with interests of their own,
apart from the interests of a dominant class (Marx and Engels 1932).10 States guard their territories
from foreign encroachment and try to maintain public order, and to these ends they raise armies and
revenues, staff bureaucracies, and gather information about their societies in order to make their task
of governing easier and more efficient. This group also recognizes the impact on domestic politics
of international forces such as war, trade, immigration, and revolutions in other countries. This
international perspective may seem natural for scholars who take an expansive, long-term, structural
view of politics, but it is not notable in the work of Barrington Moore or Gregory Luebbert, so it is
most properly a sub-group characteristic.
The earlier comparative histories tended to employ John Stuart Mill’s “methods of similarity
and difference” to test or buttress their claims. Those using Mill’s method of difference would
choose cases that had as much as possible in common except for their differing outcomes, in effect
holding the common characteristics relatively constant. These scholars could then argue that the
different outcomes would be best explained by the remaining characteristics that differed. The
method of similarity follows the opposite logic: choose cases that experienced similar outcomes
24
despite being different from one another in as many respects as possible. Using the method of
similarity, scholars argue that the similar outcomes must be attributed to the reduced set of
remaining similarities. Somewhat confusingly, Mill’s “method of difference” is the same as
Przeworski and Teune’s equally well-known “most similar systems” design, and his “method of
similarity” is identical to their “most different systems” design. Nevertheless, the logic is the same:
differences explain differences and similarities explain similarities. This was the common logic of
most comparative histories before 1990.
Since about 1987--the year of publication of Charles Ragin's The Comparative Method--
comparative histories have preferred a path-dependent model of politics, which is incompatible with
Millian comparative logic. Peter Hall has insightfully described this approach as one that sees "the
world not as a terrain marked by the operation of timeless causal regularities, but as a branching tree
whose tips represent the outcomes of events that unfold over time" (Hall 2003:385). This approach
is built around the key concepts of endogeneity and interactions. Comparative historians since 1990
have believed that outcomes occur in sequences; therefore, outcomes at one time become causes of
other outcomes at later times. This is what it means for variables to be "endogenous": they are both
causes and effects. There are interactions when causes combine to produce an outcome; and
because the effect of one cause depends on the presence of another cause, it is not useful to speak
of the independent impact of either cause separately. Recent comparative histories emphasize both
endogeneity and interactions. As Hall notes, "[t]he prototypical contention is that the impact of x
will depend on whether it occurs before or after w" (Hall 2003:385).
This is a much more complex kind of model than a simple list of variables that lead to an
outcome, and more complex than models with either endogeneity or interactions but not both.
Figure 5.3 contrasts a path-dependent model with simpler models. Model 5.3a is a basic 2-variable
25
causal model in which continuous variables A and B cause outcome Y. The same model could be
written in equation form as Y = A + B. Model 5.3b contains an interaction between A and B that
helps cause Y; or, in equation form, Y = A + B + A*B. Model 5.3c incorporates endogeneity by
distinguishing between an initial outcome Y1 and a subsequent outcome Y2. (Typically A and B
would vary with time as well.) Model 5.3c also incorporates the assumption that different variables
matter at different stages. In this case, C helps cause Y1 but not Y2 , and D helps cause Y2 but not
Y1. Representing endogeneity algebraically requires a separate equation for each outcome: in this
case, Y1 = A + B + C and Y2 =Y1 + A + B + D. Model 5.3d combines the endogeneity of Model
5.3c with the interactions of Model 5.3b. It can be written as Y1 = A + B + A*B + C and Y2 =Y1 +
A + B + A*B + D.
Model 5.3d meets Hall's criteria for path dependence, but he made his definition so general
that it includes strategic bargaining models and some complex statistical models. To narrow the
definition to the comparative histories discussed in this chapter, a further stipulation is necessary: the
models must have nominal outcomes and at least one nominal cause. To represent this feature in
Figure 5.3e, we must alter the notation. Therefore, B and Yi are no longer continuous variables, but
dichotomies. Now the upper-case "B" signifies that condition B is present, and the lower-case "b"
signifies that condition B is absent; similarly for Y1 and y1 (“not Y1”) and for Y2 and y2 (“not Y2”). In
this example, Y1 occurs only if A and B are both present in the first time period; if either A or B is
absent (signified by “a” or “b”), then the outcome is “not Y1", i.e., y1. (The branching from A to B
to Y, touching every combination of values, substitutes for the multiplicative interactions in other
models.) In the second time period, however, Y2 is determined by a different rule. If Y1 had already
been achieved, then Y2 occurs if either A or B is present; but if Y1 had not been achieved, then Y2
occurs only if both A and B are present. This argument can be expressed mathematically with
26
Boolean logic:
If [Y1 and (A or B)] or [not Y1 and A and B], then Y2;
If [Y1 and not A and not B] or [not Y1 and (not A or not B)], then not Y2.
It is the qualitative nature of the variables that produces the “branching trees” typical of path-
dependent models, as in figure 5.3e. In such models, we cannot pin down a consistent effect of A,
B, or A*B on Y2 because they may or may not be associated with Y2, depending on whether Y1 was
previously achieved or not. Figure 5.1 implies a good example of a path-dependent argument: if a
consolidated national state forms, then other conditions can eventually produce democracy there;
but if there is no consolidated national state, then these other conditions do not produce the same
democratic outcome. However, as in Figure 5.3e, outcomes need not branch away forever. In many
path-dependent models, divergent paths can later converge. In this way, path-dependent models
often identify multiple paths to a single outcome.
Evaluation of the Comparative Historical Approach
The preceding section described the nature of comparative history as an approach that is, by
varying degrees, thinner than case studies and histories but thicker than most other approaches in
comparative politics. It is also more theoretically integrated than most case studies, but only loosely
integrated in the sense of sharing some metatheoretical assumptions; and therefore less integrated
than approaches that derive propositions from a common well of formal theory or that repeatedly
test the same hypotheses with slight modifications and different samples. Comparative history also
aims for an intermediate level of generality, shunning both particularism and universalism in favor of
the middle-range aspiration of generalizing about a century or two on a continent or two. Therefore,
on all three criteria for a good theory--thickness, integration, and generality--comparative history is a
methodological compromise. Its advocates claim that this makes the approach the best of all worlds,
27
at least the best that we can actually achieve. As with any compromise, however, it is also possible
that we are left with worst of all worlds. A careful evaluation of the approach is therefore in order.
Comparative history is a very good approach for identifying and testing possible causes of
specific events in specific cases. Unfortunately, it is not as rigorous in doing this as case studies and
histories are, as they bring greater conceptual richness and much more evidence to bear on such
questions. Comparative histories do have one advantage over case studies, however. They are less
myopic: more likely to call attention to structural, macro causes.
However, comparative histories are less successful than sometimes claimed at moving
beyond specific events to develop middle-range theories. In their attempt to generalize to the middle
range, comparative histories typically encounter five problems. (1) Their concepts are slippery and
inconsistently applied. (2) They do not adequately integrate the theories developed for small
domains into a single theory for a larger domain. (3) The page limits on what presses will publish
and readers will read constrains the number of cases and the number of variables that any
comparative historian can include. (4) Comparative histories almost always end up with more
variables than cases, rendering any testing indeterminate. Finally, (5) comparative historians always
amend their theories as they go along and never test the amended theory with different evidence,
and this habit virtually guarantees capitalizing on chance. I will discuss each problem as it manifests
itself at three levels: in the typical work of comparative history, where the problems are often minor;
in the collective body of comparative historical research, where problems are most serious; and in
the potential for comparative history as an approach.
The rich conceptualization that is typical of case studies and histories becomes a burden
when scholars attempt to compare cases. It becomes impossible to respect the uniqueness of each
case and necessary to work with simpler, less precise concepts. Comparative historians must write
28
about the upper chamber of parliament instead of the British House of Lords; about the working
class instead the nitrate miners of northern Chile. Moving up "the ladder of abstraction" is a tricky
business that requires the scholar to define an umbrella concept that captures the common
characteristics of diverse times and places without ignoring other characteristics that may be causally
relevant. It is hard to do this well, so comparative historians often fall short in one of three ways.
First, some define a concept vaguely and then use it even where it does not really fit, which Sartori
(1970) called "concept stretching." Ertman, for example, cites historical research showing that earlier
comparative historians wrongly classified early modern England as a patrimonial state. In fact,
England had a more highly developed bureaucracy than Prussia did (Ertman 1997:??). Second,
some scholars define a concept in a much thinner, essentialist way but then tack on different causally
relevant qualifications in some of their cases. In this situation, the concept hides important
variations, creating an illusion of general applicability. For example, Brian Downing classified both
France and Brandenburg-Prussia as having developed "military-bureaucratic absolutism," but also
says that one reason that absolutism survived in Prussia while succumbing to revolution in France
was that France's absolutism was less absolute (Downing 1992:127-32). Third, some comparative
historians simply use vague concepts, leaving it up to the reader to guess how they should be
applied. Readers can search in vain for a concrete definition of concepts such as "a rough balance
between the crown and the nobility" (Moore), "economic dependence" (RSS), "bureaucratic state"
(Ertman), "bourgeoisie" (Moore, Skocpol), "labor-repressive agriculture" (Moore), or "solidarity of
peasant communities" (Skocpol), even though the truth of core arguments depends on how these
concepts are defined.
This problem is found even in individual works, which we should expect to be internally
consistent. There is quite a bit of variation in conceptual clarity in the genre. Ruth Collier's Paths
29
Toward Democracy (1999) stands out as unusually clear and consistent in its definitions, while Moore's
Social Origins (1966) is frustratingly muddied.11 However, the slippery or inconsistent use of concepts
is much more of problem for comparative history as a whole due to the lack of agreement among its
practitioners. One need not take an outsider's word that this problem exists. Any comparative
history selected at random will contain at least one lengthy discussion of concepts and classification
that disputes the decisions made by other comparative historians. Thus we find that Ruth Collier
classifies Weimar Germany as an episode of successful democratization while Moore, Luebbert, and
Ertman consider it a case of rising fascism (Collier 1999). Skocpol considered Moore's concept of
"labor-repressive" agriculture not to be useful and criticized Moore for using his key concept of the
"bourgeois impulse" in inconsistent ways that glossed over important differences in the strength of
the bourgeoisie (Skocpol 1973:??). And what are we to make of the fundamental disagreement
between Ruth Collier and Rueschemeyer, Stephens and Stephens? The latter argue that the working
class was the primary actor advancing the cause of democracy in Western Europe, while Collier
argues that labor organizations were not important actors in early democratization episodes in
Greece, France, Portugal, Spain, Switzerland, Norway, or Italy (Collier 1999: 23-32; Reuschemeyer,
Stephens, and Stephens 1992). The difference is a conceptual disagreement about whether the
inclusion of the working class must be accomplished before the outcome deserves to be called
"democratization." Rueschedmeyer et al. believe not, so they disqualify early steps toward
democracy that Collier considers relevant.
The second problem faced by comparative histories is the difficulty of integrating their
findings with other comparative research. This problem affects both the integration of several
comparative histories into a larger, cumulative, comparative-historical research program and the
integration of the findings from each case into a single work of comparative history. The integration
30
problem at the level of the research program is best understood as a difficulty in drawing a sample
that is comparable to the larger population of cases. If the cases selected for comparison are
comparable to all the relevant cases that could have been chosen, then one would be justified in
treating the findings of the comparison as complete, finished hypotheses ready for application to
other cases in the larger population. This is sometimes called the principle of “unit homogeneity”:
the same cause has the same effect in all the cases being analyzed (King, Keohane, and Verba
1994:91-4). But if the selected cases are systematically different from the larger population in any
relevant way, then the comparisons will suffer from selection bias. In a large sample with cases
selected at random (in a public opinion survey, for example), this is not a problem because the
sample differs only randomly, not systematically, from the population. The non-comparable aspects
of the sample can be assumed to cancel one another out, so that on average the sample is
representative of the population from which it was drawn. Unfortunately, a few dozen cases are
needed in order for the benefits of randomization to be realized, so this solution is not available to
comparative historians (King, Keohane, and Verba 1994:124-8). For them, selection bias is
unavoidable. This should be especially obvious with regard to comparative histories that focus (as
most do) on a particular geographic region or a particular period of history precisely because it is
systematically different. There is a price to pay for the advantages of middle-range theory.
Fortunately, selection bias is–in principle--correctable, but only if we know the nature of the
bias. There are several possibilities. First, the selected cases may have some fixed advantage or
disadvantage in achieving the outcome; for example, Latin American countries since the mid-1970s
tend to be more democratic than non-Latin American countries at the same level of socioeconomic
development in this period (Coppedge 2004). If we know why these countries have this bonus or
penalty, we can add an extra wrinkle to the theory to take it into account. But even if the reason is
31
unknown, it is better to know and acknowledge such systematic differences than to pretend that
they do not exist. Taking these differences into account makes sample comparable to the
population, at least in this respect. Another kind of bias occurs when the same cause has a more
powerful or a weaker effect in the selected cases than it does in other cases. A good example is the
belief that U.S. foreign policy has a more powerful impact on Central America than it does on the
Southern Cone of South America. Again, if we knew the reason for this varying effect–say, the size
and distance of each country from the U.S.–then we would have an improved theory that would
travel more easily beyond the region. Even lacking a good explanation, it is better to state that a
process is different in this region than in that one. This leaves us with a “multiple paths” kind of
theory that applies equally well to both regions.12 Finally, it is possible that the selected cases are
harder or easier to explain than the ones not selected: the outcomes just vary more. The more
apparently random variation there is in the outcomes, the more likely it is that the impact of a cause
will get lost in the noise, leading scholars to conclude, falsely, that the cause has no effect. In such
situations, the appropriate correction is to adjust the standards for each set of cases so that a big
change in the less predictable cases is treated as equivalent to a small change in the more predictable
cases. For some purposes, for example, a failed coup attempt in Northern Europe could be
equivalent (i.e., as traumatic, as severe a regime crisis) to a successful coup in Sub-Saharan Africa.
In all these situations, it would be preferable to identify explicitly how the cases selected for
a comparative history differ systematically from the rest of the world. When this is done, then we
have some guidance about how to integrate the findings of one comparative history with the
findings of other comparative histories. Failure to do so biases the conclusions of a study in
unpredictable ways, rendering effects more or less optimistic than they would be in a larger sample,
causes more or less powerful, and confidence in conclusions either inflated or deflated.
32
Without the explicit guidance needed to make the different sets of countries comparable, the
best we can do is circumscribe the relevance of the findings to just the times and places actually
studied. Although this is enough to satisfy most comparativists today, it is important to keep in mind
how far it falls short of the covering-law model of explanation, which requires universal laws. If an
explanation is limited to particular times and places for arbitrary (i.e., unspecified) reasons, then it is
an incomplete explanation, at best. Of course, it is unrealistic to expect comparative politics, still in
its infancy as a science, to produce complete explanations. Producing incomplete explanations that
work well for certain middle-range regions and periods is without doubt a useful first step along the
road to complete and therefore universal theories. In one influential view, what we learn by testing is
not so much whether a theory is true or false but “how much of the world the theory can help us
explain” (King, Keohane, and Verba 1994:101).
However, even at this early stage it is essential for comparative historians to make the
reasons for their selection decisions explicit. Otherwise, even the smallest differences in case
selection make it impossible to cumulate knowledge in a coherent research program. We have
already seen that the major comparative historical works involving democratization have used
different, though overlapping, sets of cases. This makes it all the more important for every author to
point out how the selection might have influenced the results. Downing (1992) is a good model for
others to follow. He explicitly identifies many differences between Europe and other regions:
To say that European social, political, and economic history is markedly different from that
of the rest of the world is to say nothing new. The West was the first to develop innovative
agricultural techniques, large-scale capitalist production, and a system of states. Europe was
also the first–and, unfortunately, virtually the last–to develop democratic political systems
that featured institutional checks on political monopoly, varying but frequently increasing
33
degrees of political representation, chartered rights of citizenship, and the rule of law
(Downing 1992:18).
He then offers brief case studies of Russia, Japan, and China to show that these non-western
countries “never developed constitutional government as found in late medieval Europe. Structural
configurations conducive to constitutionalism in the West, rough balance between crown and noble,
contractual-feudal military organization, and lord-peasant dynamics were absent or weak” (Downing
1992:53). Moore limits his conclusions to large, powerful states to the exclusion of smaller,
dependent states on the grounds that “the decisive causes of their politics lie outside their own
boundaries” (Moore 1966:xiii). This is brief and debatable, but at least he addressed the issue.
Many other authors discuss this issue in only the most cursory ways, if they address it at all.
The Colliers (Collier and Collier 1991:5), for example, merely note that “The present study parallels
the concerns of various analysts of Europe who have viewed the incorporation of the working class
as a pivotal transition within this larger process of societal change,” without identifying the
differences between the European and Latin American experiences. All of their discussion of case
selection deals with selection within Latin America. Skocpol (1979:40-2, 302, n97) includes a
section entitled, “Why France, Russia, and China?” but misses the point. She defends the selection
of these three cases from different regions and periods against charges that they are not comparable
to one another but neglects to address how they might not be comparable with other cases.
Elsewhere she does recognize that “there always are unexamined contextual features of the
historical cases that interact with the causes being explicitly examined in ways the comparative
historical analysis either does not reveal, or must simply assume to be irrelevant” (Skocpol 1979:39)
but does not attempt to identify these features in her analysis. Given the vague and inconsistent
justifications for case selection, we cannot assume that different comparative histories study
34
comparable sets of cases. This is an obstacle to the cumulation of findings in the comparative
historical research program.
Achieving comparability across temporal domains is as important as comparability across
spatial domains. Comparative historians should tell us not only what is systematically different about
the territories they study, but also what is different about the historical period. This kind of guidance
is necessary for unifying theories inspired by different waves of democratization. For example, it
probably matters a great deal that “first-wave” democratization in Britain took place without a
model to follow (until 1776). Democracy was not a familiar goal to achieve, but a set of evolving
institutions and practices that people invented as they went along. Transitions after World War II
had many models to emulate, and this is probably an important reason that transitions are much
faster now.
Individual works of comparative history must also demonstrate the comparability of the
cases they include. Because cases in one region during one period are relatively similar, this is usually
less difficult than ensuring the comparability of one set of cases to all other sets of cases. However,
since comparative historians are primarily interested in accounting for the differences among the
cases they have selected, they do a much better job of addressing this problem. Perhaps comparative
historians could be faulted for not accounting for the differences among their cases exhaustively. In
practice, they present a simple set of explanatory factors that most efficiently sorts the cases into the
possible outcomes. An exhaustive analysis would identify all the differences among cases and would
leave us with multiple, sometimes competing, explanations. However, this would be expecting each
comparative history to do all the work required of the entire collective comparative-historical
research program over a long period of time; it is expecting too much.
Nevertheless, contemplating the demands of rising to this challenge exposes the third
35
problem that comparative historians encounter: the practical limits to the complexity of any analysis.
Even if a scholar or teams of scholars were interested in coming closer to exhaustive explanations of
differences among a larger number of cases, no press would publish their massive writeup. And even
if a press would publish it, few readers would actually read it. There is therefore an unavoidable
tradeoff between the thickness of an analysis and the size of the domain it can cover. We can see
this tradeoff in the works discussed in this chapter. Using crude indicators of domain size–the
number of countries studied–and thickness–pages per country–we find Shaping the Political Arena at
one extreme, with 877 pages for 8 countries, or an average of 110 pages per country. (Case studies,
of course, are even thicker: several hundred book pages per country.) Skocpol and Moore are also at
the high end, with 6-8 countries and about 70 pages per country. With 11-13 cases, Luebbert and
Ertman slip to 28-38 pages per country.13 At the low extreme, Ruth Collier’s set of 22 countries and
Rueschemeyer, Stephens and Stephens’s set of 34 countries weigh in at only about 11 pages per
country. The larger the domain, the thinner the analysis.
The fourth problem of comparative history is indeterminacy. In a single-case study, many
variables are held constant over time and the only variables necessary are those that explain variation
within that case. In a comparative history, more variables are needed to account for cross-national
differences. Comparative histories face the real danger of having more variables than cases–the
“many variables, small N” problem. This is a problem because it virtually guarantees capitalizing on
chance–mistaking spurious or coincidental associations for causal relationships. The relationships
can be tested, but any tests would yield indeterminate conclusions.
If the cases had been selected at random (an important qualifier to which I will return
below), the probability that the finding of a comparative historical analysis could have been
produced by chance could be checked with Fisher’s Exact test. (The more familiar Chi-square test is
36
actually a large-sample approximation of this test. Some readers will be familiar with the Fisher test
in 2 X 2 tables. The same test, although computationally difficult, can be performed with larger
tables.) I have done this test for most of the works discussed in this chapter by summarizing their
findings in tables, which are in the appendix to this chapter. Table 5.1 presents the results of the
Fisher tests for the major comparative histories. Only two of them–Ertman and Luebbert–satisfied
the conventional p<.05 criterion for statistical significance, mostly because the other tables were
“sparsely populated,” as indicated in the “cases per cell” column. Chance results are harder to rule
out when the table is large (i.e., when there are complex explanations for complex outcomes), the
cases are few, or there are many exceptions to the pattern. Therefore, comparative histories, which
tend to have complex explanations for few cases, are prone to chance results.
Table 5.1 suggests that even if the cases were chosen at random, it would not be possible to
distinguish results this strong from sheer coincidence. But of course comparative historians do not
select their cases at random; random selection is inappropriate when the sample is small. There are
various alternative criteria for selecting cases for testing when the sample is small, some better than
others. According to King, Keohane, and Verba (1994:142), “The most egregious error is to select
observations in which the explanatory and dependent variables vary together in ways that are known
to be consistent with the hypothesis that the research purports to test.” There are two ways to do
this: selecting cases to fit the hypothesis, or selecting a hypothesis to fit the cases. I doubt that many
comparativists deliberately select cases to fit a hypothesis and then present the comparison as a test;
if they did, it would be research fraud. However, I think it is likely that some comparativists
unconsciously select cases to fit their hypothesis. The temptation to present confirming evidence
and sweep inconvenient cases under the rug is difficult to resist. Moreover, it is fine to do this as
long as such evidence is presented as illustrations of how a hypothesis might work; but this is not
37
any kind of test.
Comparative historians do, however, select hypotheses to fit their cases. In fact, they not
only do it; they admit it and praise it as methodological virtue itself. Mahoney and Rueschemeyer
(2003:13), for example, write that comparative historians can ". . .move comfortably back and forth
between theory and history in many iterations of analysis as they formulate new concepts, discover
novel explanations, and refine preexisting theoretical expectations in light of detailed case evidence."
Actually, there is methodological virtue in selecting hypotheses to fit cases, but it is a virtue for
building theory and generating hypotheses. Staying close to cases is probably the most efficient way
to develop explanations that work well for the cases being studied. But selecting hypotheses to fit
cases is not a virtue for testing. As far as testing is concerned, it is a corrupt practice. To the extent
that scholars deliberately select hypotheses to fit their cases, or unconsciously select cases to fit their
hypotheses, Fisher tests, which assume random selection, inflate the significance of the comparative
evidence for comparative historical theories. Seen in this light, the strong significance levels for the
Ertman and Luebbert models in Table 5.1 are not meaningful. Most comparative historians should
drop the pretense that there is any truly comparative testing going on in their work.
Some will object that there is plenty of testing in comparative histories. In fact, I have
already argued that comparative histories share the advantage of case studies in testing hypotheses
about the immediate causes of specific events. Designing Social Inquiry admonishes us to multiply
observable implications, to search for evidence that might disconfirm any of the many things that
must be true if the theory is true (King, Keohane, and Verba 1994:28-9), and I agree that
comparative histories have a home court advantage in this respect, even if case studies and histories
have an even greater advantage. But this is testing of hypotheses about each case. What is missing is
comparative testing of more general propositions that apply not just to one case but to all the cases in
38
the sample and to the larger population that the sample represents.
The only solution to this problem is to test the propositions that emerge from a comparative
history with a different set of cases. In principle it would be possible for a comparative historian or a
team of comparative historians to work out a model in one sample and then test its implications in a
different, sufficiently large, and fairly selected sample. A few comparative histories seem to do this,
including Moore (1966), Skocpol (1979), and Rueschemeyer et al. (1992). However, the fifth and
final problem with comparative history is that its practitioners keep amending the theory while the
"testing" is going on. If the theory building never stops, the testing cannot begin. Capitalist
Development and Democracy, for example, could be said to develop its hypotheses in Western Europe
and then test them in Latin America and the Caribbean. However, they modify their core thesis
when applying it to South America where, it turns out, the middle class sometimes played a more
important role in democratization than the working class. By making this amendment, the authors
used the South American cases to develop new theory instead of testing the theory they already had.
One of the better comparative testing efforts is Ruth Collier's Paths Toward Democracy (1999), most of
which is an out-of-sample test of the RSS thesis that the working class played a leading role in
democratization. She finds the thesis wrong in a great many cases, and this is a fine test. However,
this is the exception that proves the rule, because Collier then proceeds like other comparative
historians: she amends the hypothesis drastically by defining seven distinct patterns of
democratization, speculates about their possible causes, and never tests them all systematically or
with different cases.
The logic of comparison for testing is weak in comparative histories. Most of their value for
testing lies in the individual case studies, and even these are inferior to case studies and histories. As
a method for generating hunches and building theory, comparative history may well be the best of
39
both worlds. But with respect to testing, it is the worst of both worlds.
Nevertheless, comparative histories have made useful contributions to the study of
democratization. They have churned up relevant descriptive information, or at least brought it to the
attention of non-historians. They have developed highly plausible explanations for the unique
democratization paths taken by major countries. They have called attention to macro or structural
factors, such as class conflict, technological change, international events, and war, that case study
authors may have downplayed or taken for granted. They have cumulatively developed a framework
or meta-theory in which the material interests of rising social classes spur conflict that culminates in
the achievement of democracy. Comparative histories may have left the testing to others, but they
have played a tremendous leading role in theory building.
40
Table 5.1: Fisher Tests of Comparative-Historical Models
Author Number
of Cases
Explanatory
Categories
(rows)
Outcome
Categories
(columns)
Cases
per
Cell
Probability of
a Chance
Result
Skocpol 6 6 2 0.50 .933
The Colliers (Heritage Period) 8 4 2 0.67 .393
Rueschemeyer, Stephens and
Stephens (timing of initial
democratization in Latin
America)
10 6 4 0.42 .271
Moore 7 6 4 0.29 .188
Rueschemeyer, Stephens and
Stephens (rise of
authoritarianism in Europe)
10 6 2 0.83 .134
Downing 7 5 3 0.47 .077
Ertman (simple version) 13 4 4 0.81 .001
Luebbert 15 4 4 0.94 .000
Note: The table form of each author’s model is in the appendix to this chapter. Ruth Collier’s (1999)
model of the impact of labor on democratization considers several hypotheses but does not
integrate them into a single model. The probabilities of a chance result must be calculated for each
specific table. A number of web-based calculators will do this for small tables. I used the one at
http://www.physics.csbsju.edu/stats/exact_NROW_NCOLUMN_form.html.
41
Appendix to Chapter 5
This appendix summarizes the arguments and outcomes of the major comparative historical studiesof democratization in tabular form. Each column of a table corresponds to one of the outcomesbeing explained; each row corresponds to the combination of explanatory factors that allegedlyexplain the outome in a case or set of cases. The counts of the number of cases in each cell wereused to calculate the probabilities in Table 5.1. Fisher’s test sums the probabilities associated with allthe other possible tables with the same dimensions and the same number of cases that show astronger association between the causes and effects. Conventionally, if this sum of probabilities isless than .05, we feel confident in rejecting the possibility that the apparent association was a chanceresult.
Summary of Arguments and Outcomes in Skocpol (1979)
Causal Conditions* Social Revolution No Social Revolution N
ULBCAtrISMV France 1
UlBCAtrISMV Russia 1
ULBCatrISmv China (by 1911) 1
UlBCATRiSMvUlBCATRismv
Prussia (west of Elbe)Prussia (east of Elbe)
1
UlBcAtrIsmv Japan 1
ULbcATRIsmv England 1
N 2 4 6
*Key:Capital letters indicate the condition was present; lower-case letters indicate the condition wasabsent.U: There was a powerful landed upper class.L: The upper class had leverage over the state.B: The state was semi- or highly bureaucratic.C: The state was centralized.A: Agricultural productivity was increasing.T: There was a transition to capitalist agriculture.R: The transition to capitalist agriculture included core regions.I: International pressures were at least moderate.S: Smallholding peasants own at least 30 percent of the land.M: There is a strong peasant community.V: Villages are fairly autonomous from central control.
Source: Skocpol (1979), parts A and B of Table 1, 155-6.
42
Summary of Arguments and Outcomes about the Timing of Initial Democratization in LatinAmerica in Rueschemeyer, Stephens and Stephens (1992)
Causal Conditions None Early Medium Late N
No export expansion, labor-intensive agriculture, stateincorporation, late industrialization
Paraguay 1
Export expansion into labor-intensive agriculture, stateincorporation, medium timing ofindustrialization
Brazil 1
Export expansion into labor-intensive agriculture, incorporationby clientelistic parties, lateindustrialization
Colombia Ecuador 2
Export expansion into non- labor-intensive agriculture, incorporationby clientelistic parties, earlyindustrialization
Argentina,Uruguay
2
Mineral exports, incorporation byradical mass parties, mediumtiming of industrialization
Chile 1
Mineral exports, incorporation byradical mass parties, lateindustrialization
Peru,Venezuela
Bolivia 3
Mineral exports, revolution, earlyindustrialization
Mexico 1
N 2 3 4 2 11
Source: Author's interpretation of Rueschemeyer et al.(1992), Table 5.1, p. 164, and the surroundingtext.
Note: For the Fisher test, I excluded the row corresponding to Mexico because I could not performa Fisher test for a 7 X 4 table.
43
Summary of Arguments and Outcomes about Europe in Rueschemeyer, Stephens and Stephens (1992)
Causal Conditions*AuthoritarianBreakdown
DemocraticSurvival N
LABDr Germany, Austria 2
LAbdr Italy Australia 2
LABdr Spain 1
LAbdR United States 1
LabdR Britain 1
labdR France 1
labdr small European countries,Canada, New Zealand
3
4 6 11
Source: Rueschemeyer et al. (1992), Table 4.4, p. 144.
*Key:Capital letters indicate the condition was present; lower-case letters indicate the condition wasabsent.L: The landed upper class was politically very significant.A: The landed upper class was historically engaged in labor-repressive agriculture.B: The bourgeoisie was strong enough to be politically very significant, but not more powerful thanthe landed upper class.D: The bourgeoisie was the dependent partner in the coalition.R: There was a revolutionary break from the past.
Note: For the Fisher test, I excluded the row corresponding to the United States because I could notperform a Fisher test for a 7 X 2 table. I also treated the small European countries as a single casebecause the authors do not differentiate them.
44
Summary of the Impact of Party System Types on Coups from the Colliers (1991)
Type of Party System Coup No Coup N
Integrative Mexico, Venezuela 2
Electorally Stable butSocially Conflictual
Colombia Uruguay 2
Stalemated Peru, Argentina 2
Multi-party Polarizing Brazil, Chile 2
3 5 8
Note: The above is just one argument from the Colliers' (1991) analysis of the Heritage period, but itis the one that is most relevant for democratization. Their work advances many other argumentsabout the Heritage period and the preceding Reform, Incorporation, and Aftermath periods.
Summary of Arguments and Outcomes from Downing (1992)
Causal Conditions
Military-BureaucraticAbsolutism
Loss ofSovereignty
Preservation ofConstitu-tionalism N
High warfare, high domestic militarymobilization
Brandenburg-Prussia, France
2
High warfare, low domestic militarymobilization
Poland 1
Low warfare, low domestic militarymobilization
England to1648
1
High warfare, medium domesticmilitary mobilization aided by wealth,alliances, and geography
England 1688-1713,
Netherlands
2
High warfare, low militarymobilization but aided by access toforeign resources
Sweden 1
2 1 4 7
Source: Downing (1992), Table 2, p. 242.
45
Summary of Arguments and Outcomes in Ertman (1997)
Causal ConditionsPatrimonialAbsolutism
Bureau-cratic
Constitu-tionalism
Bureau-cratic
Absolutism
PatrimonialConstitu-tionalism N
Geopolitical competitionbefore 1450,administrative localgovernment
France, Spain,Portugal,Naples,
Tuscany,Savoy, andPapal states
7
Geopolitical competitionbefore 1450, participatorylocal government, activeparliament
Britain 1
Geopolitical competitionafter 1450, administrativelocal government
Germanstates
1
Geopolitical competitionafter 1450, participatorylocal government
Sweden Denmark Hungary andPoland
4
N 7 2 2 2 13
Source: This is the simple model presented early in the introduction to Ertman's (1997) book. By theend of the chapter, however, Ertman revised the model by distinguishing among parliaments thatwere weak (Denmark), parliaments that were initially strong but later weakened (Sweden), andparliaments that were strong and remained strong (Hungary and Poland). This change associatedeach set of causal conditions with a unique outcome. However, expanding the table from 4 X 4 to 7X 4 cells does not improve the significance of the "test."
46
Summary of Arguments and Outcomes in Luebbert (1987)
Causal ConditionsPluralist
DemocracySocial
Democracy
TraditionalDictator-
ship Fascism N
Early industrialization,dominant liberal parties,Liberal-Labor coalitions,and mass support fordemocracy
Britain,France,
Switzerland
3
Late industrialization,greater Socialist partysuccess, previouslymobilized agricultural sector
Netherlands,Belgium,Denmark,Sweden,Norway,Czecho-slovakia
6
Late industrialization,greater Socialist partysuccess, Socialistrecruitment of agriculturalproletariat, united bourgeoisparties
Austria,Finland,Hungary
3
Late industrialization,greater Socialist partysuccess, Socialistrecruitment of agriculturalproletariat, dividedbourgeois parties
Germany,Italy, Spain
3
N 3 6 3 3 15
Source: Author's synthesis of arguments in Luebbert (1986) and Luebbert (1991).
47
Summary of Arguments and Outcomes in Moore (1966)
Causal Conditions Democracy Fascism CommunismPeacefulChange N
Commercialized agriculture,strong bourgeoisie allied withreformist landed upper class,peasantry weakened
Britain 1
Commercialized agriculture,strong bourgeois impulse,peasants sweep aside landedupper class
France 1
Commercialized agriculture,strong bourgeois impulse, nopeasantry
U.S. 1
Commercialized agriculture,moderately powerfulbourgeoisie allied with landedupper class
Japan 1
Commercialized agriculture,weak bourgeoisie, landedupper class swept away bypeasant revolution
Russia, China 2
Little commercialization ofagriculture
India 1
3 1 2 1 7
Source: Because there have been many conflicting interpretations of Moore’s arguments, I relied onhis own simple summary in the preface: Moore (Moore 1966: xiv-xvii). I excluded Germany becauseMoore devotes no separate chapter to it and included India because there is a chapter on it. Chi-square results for this table would be significant if the first three rows were collapsed into one. I didnot do this because Moore takes care to differentiate the British, French, and American paths; hewould not be comfortable lumping them all together as similar cases of commercialized agriculturewith a strong bourgeois impulse, ignoring the role of the peasantry.
48
References
Allison, Graham T. 1971. Essence of decision: Explaining the Cuban Missile Crisis. Boston:
Little, Brown.
Bendix, Reinhard. 1964. Nation-Building and Citizenship: Studies of Our Changing Social
Order. New York: John Wiley and Sons.
Calabresi, Steven G. 2001. The Virtues of Presidential Government: Why Professor Ackerman is
Wrong to Prefer the German to the U.S. Constitution. Constitutional Commentary 18
(1):51-104.
Collier, David, and Robert Adcock. 1999. Democracy and Dichotomies: A Pragmatic Approach
to Choices about Concepts. Annual Review of Political Science 2:537-565.
Collier, Ruth Berins. 1999. Paths Toward Democracy: The Working Class and Elites in Western
Europe and South America. Cambridge: Cambridge University Press.
Collier, Ruth Berins, and David Collier. 1991. Shaping the Political Arena. Princenton:
Princenton University Press.
Coppedge, Michael. 2004. Explaining Democratic Deterioration in Venezuela Through Nested
Inference. In Advances and Setbacks in the Third Wave of Democracy in Latin America,
edited by F. Hagopian, and Scott Mainwaring. Cambridge: Cambridge University Press.
Diamond, Jared M. 1997. Guns, Germs, and Steel: The Fates of Human Societies. New York: W.
W. Norton.
Diamond, Martin. 1959. Democracy and the Federalist: A Reconsideration of the Framers' Intent.
American Political Science Review 53 (1):52-68.
49
Dion, Douglas. 1998. Evidence and Inference in the Comparative Case Study. Comparative
Politics 30 (2):127-145.
Downing, Brian. 1992. The Military Revolution and Political Change: Origins of Democracy
and Autocracy in Early Modern Europe. Princenton: Princenton University Press.
Eckstein, Harry. 1975. Case Study and Theory in Political Science. In Handbook of Political
Science: Strategies of Inquiry, edited by F. I. Greenstein, and Nelson W. Polsby. Reading,
Mass: Addison-Wesley.
Ertman, Thomas. 1997. Birth of the Leviathan: Building States and Regimes in Medival and
Early Modern Europe. Cambridge: Cambridge University press.
Evans, Peter B., Dietrich Rueschemeyer, and Theda Skocpol, eds. 1985. Brining the State Back
In. Cambridge: Cambridge University Press.
Fearon, James D. 1991. Counterfactuals and Hypothesis Testing in Political Science. World
Politics 43 (2):169-95.
Geddes, Barbara. 1997. Paradigms and Sandcastles: Research Design in Comparative Politics.
APSA-CP: Newsletter 8 (1):18-20.
González, Luis Eduardo, and Charles Guy Gillespie. 1994. Presidentialism and Democratic
Stability in Uruguay. In The Failure of Presidential Democracy, edited by J. J. Linz, and
Arturo Valenzuela. Baltimore: The Johns Hopkins University Press.
Hall, Peter A. 2003. Aligning Ontology and Methodology in Comparative Research. In
Comparative Historical Research in the Social Sciences, edited by J. Mahoney, and
Dietrich Rueschemeyer. Cambridge: Cambridge University Press.
Huntington, Samuel. 1993. The Clash of Civilizations. Forgein Affairs 72 (3):22-49.
50
Karl, Terry Lynn. 1997. The Paradox of Plenty: Oil Booms and Petro States. Berkeley:
University of California Press.
King, Gary, Robert Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific
Inference in Qualitative Research. Princenton: Princenton University Press.
Levine, Daniel H. 1976. Conflict and Political Change in Venezuela. Princeton: Princeton
University Press.
Lieberson, Stanley. 1992. Small Ns and Big Conclusions: An Examination of the Reasoning in
Comparative Studies Based on a Small Number of Cases. In What is a Case? Exploring
the Foundations of Social Inquiry, edited by C. C. Ragin, and Howard, S. Becker.
Cambridge: Cambridge University Press.
Luebbert, Gregory M. 1986. Comparative Democracy: Policy Making and Governing Coalitions
in Europe and Israel. New York: Columbia University Press.
Luebbert, Gregory M. 1991. Liberalism, Fascism, or Social Democracy: Social Classes and the
Political Origins of Regimes in Interwar Europe. New York: Oxford University Press.
Mahoney, James. 2003. Knowledge Accumulation in Comparative Historical Research: The Case
of Democracy and Authoritarianism. In Comparative Historical Analysis in the Social
Sciences, edited by J. Mahoney, and Dietrich Rueschemeyer. Cambridge: Cambridge
University Press.
Mahoney, James, and Dietrich Rueschemeyer. 2003. Comparative Historical Analysis:
Achievements and Agendas. In Comparative Historical Analysis in the Social Sciences,
edited by J. Mahoney, and Dietrich Rueschemeyer. Cambridge: Cambridge University
Press.
51
Malloy, James M., and Eduardo Gamarra. 1988. Revolution and Reaction: Bolivia, 1964-1985.
New Brunswick, N.J.: Transactions Bookls.
Marx, Karl, and Friedrich Engels. 1932. Manifesto of the Communist Party. New York:
International Publishers.
Moore, Barrington. 1966. Social Origins of Dictatorship and Democracy: Lord and Peasant in
the Making of the Modern World. Boston: Beacon Press.
Parry, Geraint. 1969. Political Elites, Studies in Political Science. New York: Praeger.
Popper, Karl. 1968. The Logic of Scientific Discovery. New York: Harper and Row.
Poulantzas, Nicos. 1973. Political Power and Social Classes. London: NLB Press.
Putnam, Robert D., Robert Leonardi, and Raffaela Y. Nanetti. 1993. Making Democracy Work:
Civic Transitions in Modern Italy. Princenton: Princenton University Press.
Ragin, Charles C. 1987. The Comparative Method: Beyond Qualitative and Quantitative
Strategies. Berkeley: Berkeley University Press.
Ragin, Charles C., and S. Becker Howard, eds. 1992. What is a Case? Exploring the Foundations
of Social Inquiry. Cambridge: Cambridge University Press.
Reuschemeyer, Dietrich, John D. Stephens, and Evelyne Huber Stephens. 1992. Capitalist
Development and Democracy. Chicago: University of Chicago Press.
Riggs, Fred W. 1988. The Survival of Presidentialism in America: Para-Constitutional Practices.
International Political Science Review 9 (4):247-78.
Ross, Michael L. 2001. Does Oil Hinder Democracy. World Politics 53 (3):325-61.
Sartori, Giovanni. 1970. Concept Misformation in Comparative Politics. American Political
Science Review 64 (4):1033-1053.
52
Schedler. 2003. ?????
Skocpol, Theda. 1973. A Critical Review of Barrington Moore's Social Origins of Dictatorship
and Democracy. Politics and Society 4 (1):1-34.
Skocpol, Theda. 1979. States and Social Revolutions: A Comparative Analysis of France,
Russia, and China. Cambridge: Cambridge University Press.
Stepan, Alfred. 1971. The Military in Politics: Changing Patterns in Brazil. Princenton:
Princenton University Press.
Thompson, E. P. 2001. Historical Logic [excerpt from The Poverty of Theory and Other Essays].
In The Essential E. P. Thompson, edited by D. Thompson. New York: The New Press.
Tilly, Charles, ed. 1975. The Formation of National States in Western Europe. Princeton:
Princeton University Press.
Wallerstein, Immanuel. 1974. The Modern World-System: Capitalist Agriculture and the Origins
of the European World-Economy in the Sixteenth Century. New York: Academic Press.
Webb, Robert L. 1980. Modern England: From the Eighteenth Century to the Present. 2nd ed.
New York: Harper and Row.
53
1. Some consider the definition of “a case” problematic. See Charles C. Ragin and Howard S.Becker (1992). I think that in practice and in context of a particular project, defining a case in nota problem. In the present book, a case is a country observed during a period of time.
2. Author's personal observation at an APSA meeting in the mid-1990s. As a participant on aroundtable, Linz had been critical of some attempts at theorizing. The quoted remark was hisreply to an audience member who challenged him by asking, "Should we just give up on theoryand become historians?"
3. Actually, Mom had it backwards as far as eyesight is concerned: squinting is caused by myopia(as I know well from personal experience). In comparative politics, whether a preference formicro-level explanations fosters a preference for case studies or the constraints of case studiesencourage micro-level explanations, the approach and the findings are closely associated.
4. I would not allege that the authors were “determined to blame presidentialism” if I did havepersonal knowledge of it. Both authors were advisees of Juan Linz, as I was, and all of us foundhis critique of presidentialism persuasive and saw evidence of its pernicious effects even inrelatively successful presidential democracies such as Uruguay and Venezuela. This does notmean that we were biased and wrong. Tests of the critique of presidentialism are discussed indetail in Chapter 10.
5. For an excellent summary and critique of this literature, see Michael L. Ross (2001).
6. Hypothetical reasoning is unavoidable in any kind of research, even large-sample statisticalcomparisons (Fearon 1991). In regression, we must suppose that no omitted variables arecorrelated with the variables in the model. However, this strikes me as less fanciful than thesupposition required in case studies: that there is an entirely new case that is just like theobserved case save in one respect, and that we can know the consequences of this smalldifference.
7. Their definition is useful, although they intend it to be broad enough to include statistical time-series analysis and cultural analysis, both of which I discuss in separate chapters because theyraise distinct sets of methodological issues.
8. Two important works in a similar vein are typically omitted from lists like this one: ReinhardBendix (1964) and Charles Tilly (1975). This is a bit puzzling, but it could be argued thatBendix's work is more descriptive than causal and that Tilly's book is more concerned with state-building than democratization. They tend to be classed with structural-functionalism ormodernization theory despite many arguments that are similar to those of Moore and his heirs.
9. Actually, the Colliers also include one case of social revolution–Mexico–but spend most oftheir time on the post-revolutionary period.
Endnotes
54
10. The intellectual evolution passed through a stage of holding that states are "relativelyautonomous"--sometimes acting against capital in order to guarantee the continued existence ofcapital. See Nicos Poulantzas (1973:255-321).
11. As James Mahoney observes, ". . .Moore’s specific hypotheses were stated in a relativelyvague manner that has made it difficult to for subsequent scholars to evaluate his claimsempirically. . . . [S]cholars have had difficulty summarizing Social Origins, leading to a ratherdiverse set of interpretations regarding the main arguments of the book. . . .” (Mahoney2003:137-8).
12. Actually, as noted earlier in this chapter, multiple paths are required only when the conditionon which the effect depends is conceived of in categorical terms. If it is thought of as acontinuum, then an interaction between two variables is a more precise model.
13. Here I count the Papal states and the 20 German territorial states as a single entities becauseErtman always discusses them as undifferentiated blocs. This gives him a total of 13 cases.